Table 2 Item selection methods and potential deletion criteria based on CTT and IRT
Methods | Criteria | Test aspects |
|---|---|---|
Floor effect | 20% or more individuals choose the lowest score value | Sensitivity |
Ceiling effect | 20% or more individuals choose the highest score value | Sensitivity |
t-test | The measurement variables with no statistically significant difference (α = 0.05) between 27% of the individual groups with the highest and lowest satisfaction scores | Sensitivity or discrimination |
Cronbach’s α coefficient | There is a large increase in the Cronbach’s α after the removal of a measurement variable | Internal consistency |
Item-dimension Correlation | The Pearson correlation coefficient of each measurement variable score and its corresponding latent variable score is less than 0.6 | Sensitivity or representation |
Item-total Correlation | The Pearson correlation coefficient of each measurement variable score and the model total score is less than 0.4 | Representation |
Stepwise regression | Using the total score of each latent variable as the dependent variable, the stepwise regression analysis is performed with candidate measurement variables as independent variables (sle = 0. 05, sls = 0. 10) | Importance |
Exploratory Factor analysis | Measurement variables with a factor load less than 0.4 | Representation |
a | Less than 0.3 | Discrimination |
b | Out of range of (‒4, 4) | Degree of under standability |
\(\overline{I}\) | \(\overline{I}\) Less than 0. 55 | Estimated accuracy |